-
On the vanishing of eigenfunctions of the Laplacian on tori
Authors:
Pierre Germain,
Iván Moyano,
Hui Zhu
Abstract:
Consider an eigenfunction of the Laplacian on a torus. How small can its $L^2$-norm be on small balls? We provide partial answers to this question by exploiting the distribution of integer points on spheres, basic properties of polynomials, and Nazarov--Turán type estimates for exponential polynomials. Applications to quantum limits and control theory are given.
Consider an eigenfunction of the Laplacian on a torus. How small can its $L^2$-norm be on small balls? We provide partial answers to this question by exploiting the distribution of integer points on spheres, basic properties of polynomials, and Nazarov--Turán type estimates for exponential polynomials. Applications to quantum limits and control theory are given.
△ Less
Submitted 28 June, 2024;
originally announced June 2024.
-
Phoneme Discretized Saliency Maps for Explainable Detection of AI-Generated Voice
Authors:
Shubham Gupta,
Mirco Ravanelli,
Pascal Germain,
Cem Subakan
Abstract:
In this paper, we propose Phoneme Discretized Saliency Maps (PDSM), a discretization algorithm for saliency maps that takes advantage of phoneme boundaries for explainable detection of AI-generated voice. We experimentally show with two different Text-to-Speech systems (i.e., Tacotron2 and Fastspeech2) that the proposed algorithm produces saliency maps that result in more faithful explanations com…
▽ More
In this paper, we propose Phoneme Discretized Saliency Maps (PDSM), a discretization algorithm for saliency maps that takes advantage of phoneme boundaries for explainable detection of AI-generated voice. We experimentally show with two different Text-to-Speech systems (i.e., Tacotron2 and Fastspeech2) that the proposed algorithm produces saliency maps that result in more faithful explanations compared to standard posthoc explanation methods. Moreover, by associating the saliency maps to the phoneme representations, this methodology generates explanations that tend to be more understandable than standard saliency maps on magnitude spectrograms.
△ Less
Submitted 14 June, 2024;
originally announced June 2024.
-
On the stability of the Abrikosov lattice in the Lowest Landau Level
Authors:
Pierre Germain,
Valentin Schwinte,
Laurent Thomann
Abstract:
We study the Lowest Landau Level equation set on simply and doubly-periodic domains (in other words, rectangles and strips with appropriate boundary conditions). To begin with, we study well-posedness and establish the existence of stationary solutions. Then we investigate the linear stability of the lattice solution and prove it is stable for the (hexagonal) Abrikosov lattice, but unstable for r…
▽ More
We study the Lowest Landau Level equation set on simply and doubly-periodic domains (in other words, rectangles and strips with appropriate boundary conditions). To begin with, we study well-posedness and establish the existence of stationary solutions. Then we investigate the linear stability of the lattice solution and prove it is stable for the (hexagonal) Abrikosov lattice, but unstable for rectangular lattices.
△ Less
Submitted 9 April, 2024;
originally announced April 2024.
-
Ten simple rules for collaborating with wet lab researchers for computational researchers
Authors:
Mark D. Robinson,
Peiying Cai,
Martin Emons,
Reto Gerber,
Pierre-Luc Germain,
Samuel Gunz,
Siyuan Luo,
Giulia Moro,
Emanuel Sonder,
Anthony Sonrel,
Jiayi Wang,
David Wissel,
Izaskun Mallona
Abstract:
Computational biologists are frequently engaged in collaborative data analysis with wet lab researchers. These interdisciplinary projects, as necessary as they are to the scientific endeavour, can be surprisingly challenging due to cultural differences in operations and values. In these Ten Simple Rules guide we aim to help dry lab researchers identify sources of friction; and provide actionable t…
▽ More
Computational biologists are frequently engaged in collaborative data analysis with wet lab researchers. These interdisciplinary projects, as necessary as they are to the scientific endeavour, can be surprisingly challenging due to cultural differences in operations and values. In these Ten Simple Rules guide we aim to help dry lab researchers identify sources of friction; and provide actionable tools to facilitate respectful, open, transparent and rewarding collaborations.
△ Less
Submitted 27 February, 2024;
originally announced February 2024.
-
On universality for the kinetic wave equation
Authors:
Pierre Germain,
Hui Zhu
Abstract:
On compact Riemannian manifolds with chaotic geometries, specifically those exhibiting the random wave model conjectured by Berry, we derive heuristically a homogeneous kinetic wave equation that is universal for all such manifolds.
On compact Riemannian manifolds with chaotic geometries, specifically those exhibiting the random wave model conjectured by Berry, we derive heuristically a homogeneous kinetic wave equation that is universal for all such manifolds.
△ Less
Submitted 22 February, 2024;
originally announced February 2024.
-
On the Relationship Between Interpretability and Explainability in Machine Learning
Authors:
Benjamin Leblanc,
Pascal Germain
Abstract:
Interpretability and explainability have gained more and more attention in the field of machine learning as they are crucial when it comes to high-stakes decisions and troubleshooting. Since both provide information about predictors and their decision process, they are often seen as two independent means for one single end. This view has led to a dichotomous literature: explainability techniques d…
▽ More
Interpretability and explainability have gained more and more attention in the field of machine learning as they are crucial when it comes to high-stakes decisions and troubleshooting. Since both provide information about predictors and their decision process, they are often seen as two independent means for one single end. This view has led to a dichotomous literature: explainability techniques designed for complex black-box models, or interpretable approaches ignoring the many explainability tools. In this position paper, we challenge the common idea that interpretability and explainability are substitutes for one another by listing their principal shortcomings and discussing how both of them mitigate the drawbacks of the other. In doing so, we call for a new perspective on interpretability and explainability, and works targeting both topics simultaneously, leveraging each of their respective assets.
△ Less
Submitted 25 April, 2024; v1 submitted 19 November, 2023;
originally announced November 2023.
-
Local Well-posedness for the Kinetic MMT Model
Authors:
Pierre Germain,
Joonhyun La,
Katherine Zhiyuan Zhang
Abstract:
The MMT equation was proposed by Majda, McLaughlin and Tabak as a model to study wave turbulence. We focus on the kinetic equation associated to this Hamiltonian system, which is believed to give a way to predict turbulent spectra. We clarify the formulation of the problem, and we develop the local well-posedness theory for this equation. Our analysis uncovers a surprising nonlinear smoothing phen…
▽ More
The MMT equation was proposed by Majda, McLaughlin and Tabak as a model to study wave turbulence. We focus on the kinetic equation associated to this Hamiltonian system, which is believed to give a way to predict turbulent spectra. We clarify the formulation of the problem, and we develop the local well-posedness theory for this equation. Our analysis uncovers a surprising nonlinear smoothing phenomenon.
△ Less
Submitted 18 October, 2023;
originally announced October 2023.
-
Statistical Guarantees for Variational Autoencoders using PAC-Bayesian Theory
Authors:
Sokhna Diarra Mbacke,
Florence Clerc,
Pascal Germain
Abstract:
Since their inception, Variational Autoencoders (VAEs) have become central in machine learning. Despite their widespread use, numerous questions regarding their theoretical properties remain open. Using PAC-Bayesian theory, this work develops statistical guarantees for VAEs. First, we derive the first PAC-Bayesian bound for posterior distributions conditioned on individual samples from the data-ge…
▽ More
Since their inception, Variational Autoencoders (VAEs) have become central in machine learning. Despite their widespread use, numerous questions regarding their theoretical properties remain open. Using PAC-Bayesian theory, this work develops statistical guarantees for VAEs. First, we derive the first PAC-Bayesian bound for posterior distributions conditioned on individual samples from the data-generating distribution. Then, we utilize this result to develop generalization guarantees for the VAE's reconstruction loss, as well as upper bounds on the distance between the input and the regenerated distributions. More importantly, we provide upper bounds on the Wasserstein distance between the input distribution and the distribution defined by the VAE's generative model.
△ Less
Submitted 7 December, 2023; v1 submitted 7 October, 2023;
originally announced October 2023.
-
Standing and Traveling Waves in a Nonlinearly Dispersive Lattice Model
Authors:
Ross Parker,
Pierre Germain,
Jesús Cuevas-Maraver,
Alejandro Aceves,
P. G. Kevrekidis
Abstract:
In the work of Colliander et al. (2010), a minimal lattice model was constructed describing the transfer of energy to high frequencies in the defocusing nonlinear Schrödinger equation. In the present work, we present a systematic study of the coherent structures, both standing and traveling, that arise in the context of this model. We find that the nonlinearly dispersive nature of the model is res…
▽ More
In the work of Colliander et al. (2010), a minimal lattice model was constructed describing the transfer of energy to high frequencies in the defocusing nonlinear Schrödinger equation. In the present work, we present a systematic study of the coherent structures, both standing and traveling, that arise in the context of this model. We find that the nonlinearly dispersive nature of the model is responsible for standing waves in the form of discrete compactons. On the other hand, analysis of the dynamical features of the simplest nontrivial variant of the model, namely the dimer case, yields both solutions where the intensity is trapped in a single site and solutions where the intensity moves between the two sites, which suggests the possibility of moving excitations in larger lattices. Such excitations are also suggested by the dynamical evolution associated with modulational instability. Our numerical computations confirm this expectation, and we systematically construct such traveling states as exact solutions in lattices of varying size, as well as explore their stability. A remarkable feature of these traveling lattice waves is that they are of "antidark" type, i.e., they are mounted on top of a non-vanishing background. These studies shed light on the existence, stability and dynamics of such standing and traveling states in $1+1$ dimensions, and pave the way for exploration of corresponding configurations in higher dimensions.
△ Less
Submitted 23 July, 2024; v1 submitted 20 September, 2023;
originally announced September 2023.
-
L2 to Lp bounds for spectral projectors on thin intervals in Riemannian manifolds
Authors:
Pierre Germain
Abstract:
Given a Riemannian manifold endowed with its Laplace-Beltrami operator, consider the associated spectral projector on a thin interval. As an operator from L2 to Lp, what is its operator norm? For a window of size 1, this question is fully answered by a celebrated theorem of Sogge, which applies to any manifold. For smaller windows, the global geometry of the manifold comes into play, and connectio…
▽ More
Given a Riemannian manifold endowed with its Laplace-Beltrami operator, consider the associated spectral projector on a thin interval. As an operator from L2 to Lp, what is its operator norm? For a window of size 1, this question is fully answered by a celebrated theorem of Sogge, which applies to any manifold. For smaller windows, the global geometry of the manifold comes into play, and connections to a number of mathematical fields (such as Differential Geometry, Combinatorics, Number Theory) appear, but the problem remains mostly open. The aim of this article is to review known results, focusing on cases exhibiting symmetry and emphasizing harmonic analytic rather than microlocal methods.
△ Less
Submitted 29 June, 2023;
originally announced June 2023.
-
$L^2$ to $L^p$ bounds for spectral projectors on the Euclidean two-dimensional torus
Authors:
Ciprian Demeter,
Pierre Germain
Abstract:
We consider spectral projectors associated to the Euclidean Laplacian on the two-dimensional torus, in the case where the spectral window is narrow. Bounds for their L2 to Lp operator norm are derived, extending the classical result of Sogge; a new question on the convolution kernel of the projector is introduced. The methods employed include l2 decoupling, small cap decoupling, and estimates of e…
▽ More
We consider spectral projectors associated to the Euclidean Laplacian on the two-dimensional torus, in the case where the spectral window is narrow. Bounds for their L2 to Lp operator norm are derived, extending the classical result of Sogge; a new question on the convolution kernel of the projector is introduced. The methods employed include l2 decoupling, small cap decoupling, and estimates of exponential sums.
△ Less
Submitted 30 January, 2024; v1 submitted 25 June, 2023;
originally announced June 2023.
-
Spectral projectors on hyperbolic surfaces
Authors:
Jean-Philippe Anker,
Pierre Germain,
Tristan Léger
Abstract:
In this paper, we prove $L^2 \to L^p$ estimates, where $p>2$, for spectral projectors on a wide class of hyperbolic surfaces. More precisely, we consider projections in small spectral windows $[λ-η,λ+η]$ on geometrically finite hyperbolic surfaces of infinite volume. In the convex cocompact case, we obtain optimal bounds with respect to $λ$ and $η$, up to subpolynomial losses. The proof combines t…
▽ More
In this paper, we prove $L^2 \to L^p$ estimates, where $p>2$, for spectral projectors on a wide class of hyperbolic surfaces. More precisely, we consider projections in small spectral windows $[λ-η,λ+η]$ on geometrically finite hyperbolic surfaces of infinite volume. In the convex cocompact case, we obtain optimal bounds with respect to $λ$ and $η$, up to subpolynomial losses. The proof combines the resolvent bound of Bourgain-Dyatlov and improved estimates for the Schrödinger group (Strichartz and smoothing estimates) on hyperbolic surfaces.
△ Less
Submitted 22 June, 2023;
originally announced June 2023.
-
Invariant Causal Set Covering Machines
Authors:
Thibaud Godon,
Baptiste Bauvin,
Pascal Germain,
Jacques Corbeil,
Alexandre Drouin
Abstract:
Rule-based models, such as decision trees, appeal to practitioners due to their interpretable nature. However, the learning algorithms that produce such models are often vulnerable to spurious associations and thus, they are not guaranteed to extract causally-relevant insights. In this work, we build on ideas from the invariant causal prediction literature to propose Invariant Causal Set Covering…
▽ More
Rule-based models, such as decision trees, appeal to practitioners due to their interpretable nature. However, the learning algorithms that produce such models are often vulnerable to spurious associations and thus, they are not guaranteed to extract causally-relevant insights. In this work, we build on ideas from the invariant causal prediction literature to propose Invariant Causal Set Covering Machines, an extension of the classical Set Covering Machine algorithm for conjunctions/disjunctions of binary-valued rules that provably avoids spurious associations. We demonstrate both theoretically and empirically that our method can identify the causal parents of a variable of interest in polynomial time.
△ Less
Submitted 19 July, 2023; v1 submitted 7 June, 2023;
originally announced June 2023.
-
Asymptotic Stability of Solitary Waves for One Dimensional Nonlinear Schrödinger Equations
Authors:
Charles Collot,
Pierre Germain
Abstract:
We show global asymptotic stability of solitary waves of the nonlinear Schrödinger equation in space dimension 1. Furthermore, the radiation is shown to exhibit long range scattering if the nonlinearity is cubic at the origin, or standard scattering if it is higher order. We handle a general nonlinearity without any vanishing condition, requiring that the linearized operator around the solitary wa…
▽ More
We show global asymptotic stability of solitary waves of the nonlinear Schrödinger equation in space dimension 1. Furthermore, the radiation is shown to exhibit long range scattering if the nonlinearity is cubic at the origin, or standard scattering if it is higher order. We handle a general nonlinearity without any vanishing condition, requiring that the linearized operator around the solitary wave has neither nonzero eigenvalues, nor threshold resonances. Initial data are chosen in a neighborhood of the solitary waves in the natural space $H^1 \cap L^{2,1}$ (where the latter is the weighted $L^2$ space). The proof relies on the analysis of resonances as seen through the distorted Fourier transform, combined for the first time with modulation and renormalization techniques.
△ Less
Submitted 6 June, 2023;
originally announced June 2023.
-
PAC-Bayesian Generalization Bounds for Adversarial Generative Models
Authors:
Sokhna Diarra Mbacke,
Florence Clerc,
Pascal Germain
Abstract:
We extend PAC-Bayesian theory to generative models and develop generalization bounds for models based on the Wasserstein distance and the total variation distance. Our first result on the Wasserstein distance assumes the instance space is bounded, while our second result takes advantage of dimensionality reduction. Our results naturally apply to Wasserstein GANs and Energy-Based GANs, and our boun…
▽ More
We extend PAC-Bayesian theory to generative models and develop generalization bounds for models based on the Wasserstein distance and the total variation distance. Our first result on the Wasserstein distance assumes the instance space is bounded, while our second result takes advantage of dimensionality reduction. Our results naturally apply to Wasserstein GANs and Energy-Based GANs, and our bounds provide new training objectives for these two. Although our work is mainly theoretical, we perform numerical experiments showing non-vacuous generalization bounds for Wasserstein GANs on synthetic datasets.
△ Less
Submitted 13 November, 2023; v1 submitted 17 February, 2023;
originally announced February 2023.
-
Seeking Interpretability and Explainability in Binary Activated Neural Networks
Authors:
Benjamin Leblanc,
Pascal Germain
Abstract:
We study the use of binary activated neural networks as interpretable and explainable predictors in the context of regression tasks on tabular data; more specifically, we provide guarantees on their expressiveness, present an approach based on the efficient computation of SHAP values for quantifying the relative importance of the features, hidden neurons and even weights. As the model's simplicity…
▽ More
We study the use of binary activated neural networks as interpretable and explainable predictors in the context of regression tasks on tabular data; more specifically, we provide guarantees on their expressiveness, present an approach based on the efficient computation of SHAP values for quantifying the relative importance of the features, hidden neurons and even weights. As the model's simplicity is instrumental in achieving interpretability, we propose a greedy algorithm for building compact binary activated networks. This approach doesn't need to fix an architecture for the network in advance: it is built one layer at a time, one neuron at a time, leading to predictors that aren't needlessly complex for a given task.
△ Less
Submitted 10 June, 2024; v1 submitted 7 September, 2022;
originally announced September 2022.
-
Stability and cascades for the Kolmogorov-Zakharov spectrum of wave turbulence
Authors:
Charles Collot,
Helge Dietert,
Pierre Germain
Abstract:
The kinetic wave equation arises in wave turbulence to describe the Fourier spectrum of solutions to the cubic Schroedinger equation. The equation has two Kolmogorov-Zakharov steady states corresponding to out-of-equilibrium cascades transferring for the first solution mass from infinity to zero (small spatial scales to large scales) and for the other solution energy from zero to infinity. After c…
▽ More
The kinetic wave equation arises in wave turbulence to describe the Fourier spectrum of solutions to the cubic Schroedinger equation. The equation has two Kolmogorov-Zakharov steady states corresponding to out-of-equilibrium cascades transferring for the first solution mass from infinity to zero (small spatial scales to large scales) and for the other solution energy from zero to infinity. After conjecturing the generic development of the two cascades, we verify it partially in the isotropic case by proving the nonlinear stability of the mass cascade in the stationary setting. This constructs non-trivial out-of-equilibrium steady states with a direct energy cascade as well as an indirect mass cascade.
△ Less
Submitted 12 September, 2022; v1 submitted 1 August, 2022;
originally announced August 2022.
-
Bounds for spectral projectors on generic tori
Authors:
Pierre Germain,
Simon L. Rydin Myerson
Abstract:
We investigate norms of spectral projectors on thin spherical shells for the Laplacian on generic tori, including generic rectangular tori. We state a conjecture and partially prove it, improving on previous results concerning arbitrary tori.
We investigate norms of spectral projectors on thin spherical shells for the Laplacian on generic tori, including generic rectangular tori. We state a conjecture and partially prove it, improving on previous results concerning arbitrary tori.
△ Less
Submitted 2 December, 2022; v1 submitted 6 July, 2022;
originally announced July 2022.
-
On the derivation of the homogeneous kinetic wave equation for a nonlinear random matrix model
Authors:
Guillaume Dubach,
Pierre Germain,
Benjamin Harrop-Griffiths
Abstract:
We consider a nonlinear system of ODEs, where the underlying linear dynamics are determined by a Hermitian random matrix ensemble. We prove that the leading order dynamics in the weakly nonlinear, infinite volume limit are determined by a solution to the corresponding kinetic wave equation on a non-trivial timescale. Our proof relies on estimates for Haar-distributed unitary matrices obtained from…
▽ More
We consider a nonlinear system of ODEs, where the underlying linear dynamics are determined by a Hermitian random matrix ensemble. We prove that the leading order dynamics in the weakly nonlinear, infinite volume limit are determined by a solution to the corresponding kinetic wave equation on a non-trivial timescale. Our proof relies on estimates for Haar-distributed unitary matrices obtained from Weingarten calculus, which may be of independent interest.
△ Less
Submitted 25 September, 2023; v1 submitted 25 March, 2022;
originally announced March 2022.
-
Bounds for spectral projectors on the Euclidean cylinder
Authors:
Pierre Germain,
Simon L. Rydin Myerson
Abstract:
We prove essentially optimal bounds for norms of spectral projectors on thin spherical shells for the Laplacian on the cylinder (R/Z)*R. In contrast to previous investigations into spectral projectors on tori, having one unbounded dimension available permits a compact self-contained proof.
We prove essentially optimal bounds for norms of spectral projectors on thin spherical shells for the Laplacian on the cylinder (R/Z)*R. In contrast to previous investigations into spectral projectors on tori, having one unbounded dimension available permits a compact self-contained proof.
△ Less
Submitted 15 March, 2022;
originally announced March 2022.
-
On $1$d quadratic Klein-Gordon equations with a potential and symmetries
Authors:
Pierre Germain,
Fabio Pusateri,
Katherine Zhiyuan Zhang
Abstract:
This paper is a continuation of a previous work Germain-Pusateri (2020) by the first two authors. We focus on $1$ dimensional quadratic Klein-Gordon equations with a potential, under some assumptions that are less general than Germain-Pusateri (2020), but allow us to present some simplifications in the proof of global existence with decay for small solutions. In particular, we can propagate a stro…
▽ More
This paper is a continuation of a previous work Germain-Pusateri (2020) by the first two authors. We focus on $1$ dimensional quadratic Klein-Gordon equations with a potential, under some assumptions that are less general than Germain-Pusateri (2020), but allow us to present some simplifications in the proof of global existence with decay for small solutions. In particular, we can propagate a stronger control on a basic $L^2$-weighted type norm while providing some shorter and less technical proofs for some of the arguments.
△ Less
Submitted 1 February, 2023; v1 submitted 26 February, 2022;
originally announced February 2022.
-
PAC-Bayesian Learning of Aggregated Binary Activated Neural Networks with Probabilities over Representations
Authors:
Louis Fortier-Dubois,
Gaël Letarte,
Benjamin Leblanc,
François Laviolette,
Pascal Germain
Abstract:
Considering a probability distribution over parameters is known as an efficient strategy to learn a neural network with non-differentiable activation functions. We study the expectation of a probabilistic neural network as a predictor by itself, focusing on the aggregation of binary activated neural networks with normal distributions over real-valued weights. Our work leverages a recent analysis d…
▽ More
Considering a probability distribution over parameters is known as an efficient strategy to learn a neural network with non-differentiable activation functions. We study the expectation of a probabilistic neural network as a predictor by itself, focusing on the aggregation of binary activated neural networks with normal distributions over real-valued weights. Our work leverages a recent analysis derived from the PAC-Bayesian framework that derives tight generalization bounds and learning procedures for the expected output value of such an aggregation, which is given by an analytical expression. While the combinatorial nature of the latter has been circumvented by approximations in previous works, we show that the exact computation remains tractable for deep but narrow neural networks, thanks to a dynamic programming approach. This leads us to a peculiar bound minimization learning algorithm for binary activated neural networks, where the forward pass propagates probabilities over representations instead of activation values. A stochastic counterpart that scales to wide architectures is proposed.
△ Less
Submitted 14 April, 2023; v1 submitted 28 October, 2021;
originally announced October 2021.
-
Stratospheric planetary flows from the perspective of the Euler equation on a rotating sphere
Authors:
Adrian Constantin,
Pierre Germain
Abstract:
This article is devoted to stationary solutions of Euler's equation on a rotating sphere, and to their relevance to the dynamics of stratospheric flows in the atmosphere of the outer planets of our solar system and in polar regions of the Earth. For the Euler equation, under appropriate conditions, rigidity results are established, ensuring that the solutions are either zonal or rotated zonal solu…
▽ More
This article is devoted to stationary solutions of Euler's equation on a rotating sphere, and to their relevance to the dynamics of stratospheric flows in the atmosphere of the outer planets of our solar system and in polar regions of the Earth. For the Euler equation, under appropriate conditions, rigidity results are established, ensuring that the solutions are either zonal or rotated zonal solutions. A natural analogue of Arnold's stability criterion is proved. In both cases, the lowest mode Rossby-Haurwitz stationary solutions (more precisely, those whose stream functions belong to the sum of the first two eigenspaces of the Laplace-Beltrami operator) appear as limiting cases. We study the stability properties of these critical stationary solutions. Results on the local and global bifurcation of non-zonal stationary solutions from classical Rossby-Haurwitz waves are also obtained. Finally, we show that stationary solutions of the Euler equation on a rotating sphere are building blocks for travelling-wave solutions of the 3D system that describes the leading order dynamics of stratospheric planetary flows, capturing the characteristic decrease of density and increase of temperature with height in this region of the atmosphere.
△ Less
Submitted 17 September, 2021;
originally announced September 2021.
-
Derivation of the kinetic wave equation for quadratic dispersive problems in the inhomogeneous setting
Authors:
Ioakeim Ampatzoglou,
Charles Collot,
Pierre Germain
Abstract:
We examine the validity of the kinetic description of wave turbulence for a model quadratic equation. We focus on the space-inhomogeneous case, which had not been treated earlier; the space-homogeneous case is a simple variant. We determine nonlinearities for which the kinetic description holds, or might fail, up to an arbitrarily small polynomial loss of the kinetic time scale. More precisely, we…
▽ More
We examine the validity of the kinetic description of wave turbulence for a model quadratic equation. We focus on the space-inhomogeneous case, which had not been treated earlier; the space-homogeneous case is a simple variant. We determine nonlinearities for which the kinetic description holds, or might fail, up to an arbitrarily small polynomial loss of the kinetic time scale. More precisely, we focus on the convergence of the Dyson series, which is an expansion of the solution in terms of the random data.
△ Less
Submitted 2 May, 2024; v1 submitted 25 July, 2021;
originally announced July 2021.
-
Learning Stochastic Majority Votes by Minimizing a PAC-Bayes Generalization Bound
Authors:
Valentina Zantedeschi,
Paul Viallard,
Emilie Morvant,
Rémi Emonet,
Amaury Habrard,
Pascal Germain,
Benjamin Guedj
Abstract:
We investigate a stochastic counterpart of majority votes over finite ensembles of classifiers, and study its generalization properties. While our approach holds for arbitrary distributions, we instantiate it with Dirichlet distributions: this allows for a closed-form and differentiable expression for the expected risk, which then turns the generalization bound into a tractable training objective.…
▽ More
We investigate a stochastic counterpart of majority votes over finite ensembles of classifiers, and study its generalization properties. While our approach holds for arbitrary distributions, we instantiate it with Dirichlet distributions: this allows for a closed-form and differentiable expression for the expected risk, which then turns the generalization bound into a tractable training objective. The resulting stochastic majority vote learning algorithm achieves state-of-the-art accuracy and benefits from (non-vacuous) tight generalization bounds, in a series of numerical experiments when compared to competing algorithms which also minimize PAC-Bayes objectives -- both with uninformed (data-independent) and informed (data-dependent) priors.
△ Less
Submitted 19 October, 2021; v1 submitted 23 June, 2021;
originally announced June 2021.
-
Self-Bounding Majority Vote Learning Algorithms by the Direct Minimization of a Tight PAC-Bayesian C-Bound
Authors:
Paul Viallard,
Pascal Germain,
Amaury Habrard,
Emilie Morvant
Abstract:
In the PAC-Bayesian literature, the C-Bound refers to an insightful relation between the risk of a majority vote classifier (under the zero-one loss) and the first two moments of its margin (i.e., the expected margin and the voters' diversity). Until now, learning algorithms developed in this framework minimize the empirical version of the C-Bound, instead of explicit PAC-Bayesian generalization b…
▽ More
In the PAC-Bayesian literature, the C-Bound refers to an insightful relation between the risk of a majority vote classifier (under the zero-one loss) and the first two moments of its margin (i.e., the expected margin and the voters' diversity). Until now, learning algorithms developed in this framework minimize the empirical version of the C-Bound, instead of explicit PAC-Bayesian generalization bounds. In this paper, by directly optimizing PAC-Bayesian guarantees on the C-Bound, we derive self-bounding majority vote learning algorithms. Moreover, our algorithms based on gradient descent are scalable and lead to accurate predictors paired with non-vacuous guarantees.
△ Less
Submitted 31 August, 2021; v1 submitted 28 April, 2021;
originally announced April 2021.
-
Bounds for spectral projectors on tori
Authors:
Pierre Germain,
Simon Leo Rydin Myerson
Abstract:
We investigate norms of spectral projectors on thin spherical shells for the Laplacian on tori. This is closely related to the boundedness of resolvents of the Laplacian, and to the boundedness of $L^p$ norms of eigenfunctions of the Laplacian. We formulate a conjecture, and partially prove it.
We investigate norms of spectral projectors on thin spherical shells for the Laplacian on tori. This is closely related to the boundedness of resolvents of the Laplacian, and to the boundedness of $L^p$ norms of eigenfunctions of the Laplacian. We formulate a conjecture, and partially prove it.
△ Less
Submitted 22 March, 2022; v1 submitted 27 April, 2021;
originally announced April 2021.
-
Spectral projectors, resolvent, and Fourier restriction on the hyperbolic space
Authors:
Pierre Germain,
Tristan Léger
Abstract:
We develop a unified approach to proving $L^p-L^q$ boundedness of spectral projectors, the resolvent of the Laplace-Beltrami operator and its derivative on $\mathbb{H}^d.$ In the case of spectral projectors, and when $p$ and $q$ are in duality, the dependence of the implicit constant on $p$ is shown to be sharp. We also give partial results on the question of $L^p-L^q$ boundedness of the Fourier e…
▽ More
We develop a unified approach to proving $L^p-L^q$ boundedness of spectral projectors, the resolvent of the Laplace-Beltrami operator and its derivative on $\mathbb{H}^d.$ In the case of spectral projectors, and when $p$ and $q$ are in duality, the dependence of the implicit constant on $p$ is shown to be sharp. We also give partial results on the question of $L^p-L^q$ boundedness of the Fourier extension operator. As an application, we prove smoothing estimates for the free Schrödinger equation on $\mathbb{H}^d$ and a limiting absorption principle for the electromagnetic Schrödinger equation with small potentials.
△ Less
Submitted 21 June, 2023; v1 submitted 8 April, 2021;
originally announced April 2021.
-
A General Framework for the Practical Disintegration of PAC-Bayesian Bounds
Authors:
Paul Viallard,
Pascal Germain,
Amaury Habrard,
Emilie Morvant
Abstract:
PAC-Bayesian bounds are known to be tight and informative when studying the generalization ability of randomized classifiers. However, they require a loose and costly derandomization step when applied to some families of deterministic models such as neural networks. As an alternative to this step, we introduce new PAC-Bayesian generalization bounds that have the originality to provide disintegrate…
▽ More
PAC-Bayesian bounds are known to be tight and informative when studying the generalization ability of randomized classifiers. However, they require a loose and costly derandomization step when applied to some families of deterministic models such as neural networks. As an alternative to this step, we introduce new PAC-Bayesian generalization bounds that have the originality to provide disintegrated bounds, i.e., they give guarantees over one single hypothesis instead of the usual averaged analysis. Our bounds are easily optimizable and can be used to design learning algorithms. We illustrate this behavior on neural networks, and we show a significant practical improvement over the state-of-the-art framework.
△ Less
Submitted 18 September, 2023; v1 submitted 17 February, 2021;
originally announced February 2021.
-
Out-of-distribution detection for regression tasks: parameter versus predictor entropy
Authors:
Yann Pequignot,
Mathieu Alain,
Patrick Dallaire,
Alireza Yeganehparast,
Pascal Germain,
Josée Desharnais,
François Laviolette
Abstract:
It is crucial to detect when an instance lies downright too far from the training samples for the machine learning model to be trusted, a challenge known as out-of-distribution (OOD) detection. For neural networks, one approach to this task consists of learning a diversity of predictors that all can explain the training data. This information can be used to estimate the epistemic uncertainty at a…
▽ More
It is crucial to detect when an instance lies downright too far from the training samples for the machine learning model to be trusted, a challenge known as out-of-distribution (OOD) detection. For neural networks, one approach to this task consists of learning a diversity of predictors that all can explain the training data. This information can be used to estimate the epistemic uncertainty at a given newly observed instance in terms of a measure of the disagreement of the predictions. Evaluation and certification of the ability of a method to detect OOD require specifying instances which are likely to occur in deployment yet on which no prediction is available. Focusing on regression tasks, we choose a simple yet insightful model for this OOD distribution and conduct an empirical evaluation of the ability of various methods to discriminate OOD samples from the data. Moreover, we exhibit evidence that a diversity of parameters may fail to translate to a diversity of predictors. Based on the choice of an OOD distribution, we propose a new way of estimating the entropy of a distribution on predictors based on nearest neighbors in function space. This leads to a variational objective which, combined with the family of distributions given by a generative neural network, systematically produces a diversity of predictors that provides a robust way to detect OOD samples.
△ Less
Submitted 11 September, 2023; v1 submitted 24 October, 2020;
originally announced October 2020.
-
Derivation of the homogeneous kinetic wave equation: longer time scales
Authors:
Charles Collot,
Pierre Germain
Abstract:
We consider the nonlinear Schrödinger equation set on a flat torus, in the regime which is conjectured to lead to the kinetic wave equation; in particular, the data are random, and spread up to high frequency in a weakly nonlinear regime. We pursue the investigations of our previous paper, and show that, in the case where the torus is the standard one, only the scaling considered there allows conv…
▽ More
We consider the nonlinear Schrödinger equation set on a flat torus, in the regime which is conjectured to lead to the kinetic wave equation; in particular, the data are random, and spread up to high frequency in a weakly nonlinear regime. We pursue the investigations of our previous paper, and show that, in the case where the torus is the standard one, only the scaling considered there allows convergence of the Dyson series up to the kinetic time scale. We also show that, for generic quadratic dispersion relations (non rectangular tori), the Dyson series converges on significantly longer time scales; we are able to reach the kinetic time up to an arbitrarily small polynomial error for a larger set of scalings. These results show the importance of the exact structure of the dispersion relation, more specifically of equidistribution properties of some bilinear quantities akin to pair correlations derived from it.
△ Less
Submitted 7 July, 2020;
originally announced July 2020.
-
Quadratic Klein-Gordon equations with a potential in one dimension
Authors:
Pierre Germain,
Fabio Pusateri
Abstract:
This paper proposes a fairly general new point of view on the question of asymptotic stability of (topological) solitons. Our approach is based on the use of the distorted Fourier transform at the nonlinear level; it does not rely on Strichartz or virial estimates and is therefore able to treat low power nonlinearities (hence also non-localized solitons) and capture the global (in space and time)…
▽ More
This paper proposes a fairly general new point of view on the question of asymptotic stability of (topological) solitons. Our approach is based on the use of the distorted Fourier transform at the nonlinear level; it does not rely on Strichartz or virial estimates and is therefore able to treat low power nonlinearities (hence also non-localized solitons) and capture the global (in space and time) behavior of solutions. More specifically, we consider quadratic nonlinear Klein-Gordon equations with a potential in one space dimension. The potential is assumed to be regular, decaying, and either generic or exceptional (with some additional parity assumptions). Assuming that the associated Schrödinger operator has no negative eigenvalues, we obtain global-in-time bounds, including sharp pointwise decay and modified asymptotics, for small solutions. These results have implications for the asymptotic stability of solitons, or topological solitons, for a variety of problems. For instance, we obtain full asymptotic stability of kinks with respect to odd perturbations for the double Sine-Gordon problem (in an appropriate range of the deformation parameter). For the $φ^4$ problem, we obtain asymptotic stability of the kink (with respect to odd perturbations) when the coupling to the internal mode appearing in the linearization around it is neglected. Our results also go beyond these examples since our approach allows for the presence of a fully coherent phenomenon at the level of quadratic interactions, which creates a degeneracy in distorted Fourier space. We devise a suitable framework that incorporates this, and use multilinear harmonic analysis in the distorted setting to control all nonlinear interactions.
△ Less
Submitted 14 February, 2023; v1 submitted 28 June, 2020;
originally announced June 2020.
-
On the derivation of the homogeneous kinetic wave equation
Authors:
Charles Collot,
Pierre Germain
Abstract:
The nonlinear Schrödinger equation in the weakly nonlinear regime with random Gaussian fields as initial data is considered. The problem is set on the torus in any dimension greater than two. A conjecture in statistical physics is that there exists a kinetic time scale depending on the frequency localisation of the data and on the strength of the nonlinearity, on which the expectation of the squar…
▽ More
The nonlinear Schrödinger equation in the weakly nonlinear regime with random Gaussian fields as initial data is considered. The problem is set on the torus in any dimension greater than two. A conjecture in statistical physics is that there exists a kinetic time scale depending on the frequency localisation of the data and on the strength of the nonlinearity, on which the expectation of the squares of moduli of Fourier modes evolve according to an effective equation: the so-called kinetic wave equation. When the kinetic time for our setup is $1$, we prove this conjecture up to an arbitrarily small polynomial loss. When the kinetic time is larger than $1$, we obtain its validity on a more restricted time scale. The key idea of the proof is the use of Feynman interaction diagrams both in the construction of an approximate solution and in the study of its nonlinear stability. We perform a truncated series expansion in the initial data, and obtain bounds in average in various function spaces for its elements. The linearised dynamics then involves a linear Schrödinger equation with a corresponding random potential. We bound the expectation of the operator norm in Bourgain spaces using diagrams and random matrix tools. This gives a new approach for the analysis of nonlinear wave equations out of equilibrium, and gives hope that refinements of the method could help settle the conjecture.
△ Less
Submitted 18 February, 2021; v1 submitted 21 December, 2019;
originally announced December 2019.
-
Improved PAC-Bayesian Bounds for Linear Regression
Authors:
Vera Shalaeva,
Alireza Fakhrizadeh Esfahani,
Pascal Germain,
Mihaly Petreczky
Abstract:
In this paper, we improve the PAC-Bayesian error bound for linear regression derived in Germain et al. [10]. The improvements are twofold. First, the proposed error bound is tighter, and converges to the generalization loss with a well-chosen temperature parameter. Second, the error bound also holds for training data that are not independently sampled. In particular, the error bound applies to cer…
▽ More
In this paper, we improve the PAC-Bayesian error bound for linear regression derived in Germain et al. [10]. The improvements are twofold. First, the proposed error bound is tighter, and converges to the generalization loss with a well-chosen temperature parameter. Second, the error bound also holds for training data that are not independently sampled. In particular, the error bound applies to certain time series generated by well-known classes of dynamical models, such as ARX models.
△ Less
Submitted 6 December, 2019;
originally announced December 2019.
-
On self-similar solutions to degenerate compressible Navier-Stokes equations
Authors:
Pierre Germain,
Tsukasa Iwabuchi,
Tristan Léger
Abstract:
We study cavitating self-similar solutions to compressible Navier-Stokes equations with degenerate density-dependent viscosity. We prove both existence of expanders and non-existence of small shrinkers.
We study cavitating self-similar solutions to compressible Navier-Stokes equations with degenerate density-dependent viscosity. We prove both existence of expanders and non-existence of small shrinkers.
△ Less
Submitted 8 November, 2019;
originally announced November 2019.
-
Backward self-similar solutions for compressible Navier-Stokes equations
Authors:
Pierre Germain,
Tsukasa Iwabuchi,
Tristan Léger
Abstract:
This article is devoted to backward self-similar blow up solutions of the compressible Navier-Stokes equations with radial symmetry. We show that such solutions cannot exist if they either satisfy an appropriate smallness condition, or have finite energy. Furthermore, numerical simulations lead us to the conjecture that such solutions do not exist.
This article is devoted to backward self-similar blow up solutions of the compressible Navier-Stokes equations with radial symmetry. We show that such solutions cannot exist if they either satisfy an appropriate smallness condition, or have finite energy. Furthermore, numerical simulations lead us to the conjecture that such solutions do not exist.
△ Less
Submitted 31 October, 2019;
originally announced November 2019.
-
PAC-Bayesian Contrastive Unsupervised Representation Learning
Authors:
Kento Nozawa,
Pascal Germain,
Benjamin Guedj
Abstract:
Contrastive unsupervised representation learning (CURL) is the state-of-the-art technique to learn representations (as a set of features) from unlabelled data. While CURL has collected several empirical successes recently, theoretical understanding of its performance was still missing. In a recent work, Arora et al. (2019) provide the first generalisation bounds for CURL, relying on a Rademacher c…
▽ More
Contrastive unsupervised representation learning (CURL) is the state-of-the-art technique to learn representations (as a set of features) from unlabelled data. While CURL has collected several empirical successes recently, theoretical understanding of its performance was still missing. In a recent work, Arora et al. (2019) provide the first generalisation bounds for CURL, relying on a Rademacher complexity. We extend their framework to the flexible PAC-Bayes setting, allowing us to deal with the non-iid setting. We present PAC-Bayesian generalisation bounds for CURL, which are then used to derive a new representation learning algorithm. Numerical experiments on real-life datasets illustrate that our algorithm achieves competitive accuracy, and yields non-vacuous generalisation bounds.
△ Less
Submitted 17 July, 2020; v1 submitted 10 October, 2019;
originally announced October 2019.
-
Onset of the wave turbulence description of the longtime behavior of the nonlinear Schrödinger equation
Authors:
Tristan Buckmaster,
Pierre Germain,
Zaher Hani,
Jalal Shatah
Abstract:
Consider the cubic nonlinear Schrödinger equation set on a d-dimensional torus, with data whose Fourier coefficients have phases which are uniformly distributed and independent. We show that, on average, the evolution of the moduli of the Fourier coefficients is governed by the so-called wave kinetic equation, predicted in wave turbulence theory, on a nontrivial timescale.
Consider the cubic nonlinear Schrödinger equation set on a d-dimensional torus, with data whose Fourier coefficients have phases which are uniformly distributed and independent. We show that, on average, the evolution of the moduli of the Fourier coefficients is governed by the so-called wave kinetic equation, predicted in wave turbulence theory, on a nontrivial timescale.
△ Less
Submitted 12 March, 2021; v1 submitted 8 July, 2019;
originally announced July 2019.
-
Learning Landmark-Based Ensembles with Random Fourier Features and Gradient Boosting
Authors:
Léo Gautheron,
Pascal Germain,
Amaury Habrard,
Emilie Morvant,
Marc Sebban,
Valentina Zantedeschi
Abstract:
We propose a Gradient Boosting algorithm for learning an ensemble of kernel functions adapted to the task at hand. Unlike state-of-the-art Multiple Kernel Learning techniques that make use of a pre-computed dictionary of kernel functions to select from, at each iteration we fit a kernel by approximating it as a weighted sum of Random Fourier Features (RFF) and by optimizing their barycenter. This…
▽ More
We propose a Gradient Boosting algorithm for learning an ensemble of kernel functions adapted to the task at hand. Unlike state-of-the-art Multiple Kernel Learning techniques that make use of a pre-computed dictionary of kernel functions to select from, at each iteration we fit a kernel by approximating it as a weighted sum of Random Fourier Features (RFF) and by optimizing their barycenter. This allows us to obtain a more versatile method, easier to setup and likely to have better performance. Our study builds on a recent result showing one can learn a kernel from RFF by computing the minimum of a PAC-Bayesian bound on the kernel alignment generalization loss, which is obtained efficiently from a closed-form solution. We conduct an experimental analysis to highlight the advantages of our method w.r.t. both Boosting-based and kernel-learning state-of-the-art methods.
△ Less
Submitted 14 June, 2019;
originally announced June 2019.
-
Dichotomize and Generalize: PAC-Bayesian Binary Activated Deep Neural Networks
Authors:
Gaël Letarte,
Pascal Germain,
Benjamin Guedj,
François Laviolette
Abstract:
We present a comprehensive study of multilayer neural networks with binary activation, relying on the PAC-Bayesian theory. Our contributions are twofold: (i) we develop an end-to-end framework to train a binary activated deep neural network, (ii) we provide nonvacuous PAC-Bayesian generalization bounds for binary activated deep neural networks. Our results are obtained by minimizing the expected l…
▽ More
We present a comprehensive study of multilayer neural networks with binary activation, relying on the PAC-Bayesian theory. Our contributions are twofold: (i) we develop an end-to-end framework to train a binary activated deep neural network, (ii) we provide nonvacuous PAC-Bayesian generalization bounds for binary activated deep neural networks. Our results are obtained by minimizing the expected loss of an architecture-dependent aggregation of binary activated deep neural networks. Our analysis inherently overcomes the fact that binary activation function is non-differentiable. The performance of our approach is assessed on a thorough numerical experiment protocol on real-life datasets.
△ Less
Submitted 4 February, 2020; v1 submitted 24 May, 2019;
originally announced May 2019.
-
Self-similar solutions for compressible Navier-Stokes equations
Authors:
Pierre Germain,
Tsukasa Iwabuchi
Abstract:
We construct forward self-similar solutions (expanders) for the compressible Navier-Stokes equations. Some of these self-similar solutions are smooth, while others exhibit a singularity do to cavitation at the origin.
We construct forward self-similar solutions (expanders) for the compressible Navier-Stokes equations. Some of these self-similar solutions are smooth, while others exhibit a singularity do to cavitation at the origin.
△ Less
Submitted 24 March, 2019;
originally announced March 2019.
-
Pseudo-Bayesian Learning with Kernel Fourier Transform as Prior
Authors:
Gaël Letarte,
Emilie Morvant,
Pascal Germain
Abstract:
We revisit Rahimi and Recht (2007)'s kernel random Fourier features (RFF) method through the lens of the PAC-Bayesian theory. While the primary goal of RFF is to approximate a kernel, we look at the Fourier transform as a prior distribution over trigonometric hypotheses. It naturally suggests learning a posterior on these hypotheses. We derive generalization bounds that are optimized by learning a…
▽ More
We revisit Rahimi and Recht (2007)'s kernel random Fourier features (RFF) method through the lens of the PAC-Bayesian theory. While the primary goal of RFF is to approximate a kernel, we look at the Fourier transform as a prior distribution over trigonometric hypotheses. It naturally suggests learning a posterior on these hypotheses. We derive generalization bounds that are optimized by learning a pseudo-posterior obtained from a closed-form expression. Based on this study, we consider two learning strategies: The first one finds a compact landmarks-based representation of the data where each landmark is given by a distribution-tailored similarity measure, while the second one provides a PAC-Bayesian justification to the kernel alignment method of Sinha and Duchi (2016).
△ Less
Submitted 27 March, 2019; v1 submitted 30 October, 2018;
originally announced October 2018.
-
Strichartz estimates for the Schroedinger equation on non-rectangular two-dimensional tori
Authors:
Yu Deng,
Pierre Germain,
Larry Guth,
Simon Myerson
Abstract:
We propose a conjecture for long time Strichartz estimates on generic (non-rectangular) flat tori. We proceed to partially prove it in dimension 2. Our arguments involve on the one hand Weyl bounds; and on the other hands bounds on the number of solutions of Diophantine problems.
We propose a conjecture for long time Strichartz estimates on generic (non-rectangular) flat tori. We proceed to partially prove it in dimension 2. Our arguments involve on the one hand Weyl bounds; and on the other hands bounds on the number of solutions of Diophantine problems.
△ Less
Submitted 28 April, 2021; v1 submitted 12 October, 2018;
originally announced October 2018.
-
Vortex filament solutions of the Navier-Stokes equations
Authors:
Jacob Bedrossian,
Pierre Germain,
Benjamin Harrop-Griffiths
Abstract:
We consider solutions of the Navier-Stokes equations in $3d$ with vortex filament initial data of arbitrary circulation, that is, initial vorticity given by a divergence-free vector-valued measure of arbitrary mass supported on a smooth curve. First, we prove global well-posedness for perturbations of the Oseen vortex column in scaling-critical spaces. Second, we prove local well-posedness (in a s…
▽ More
We consider solutions of the Navier-Stokes equations in $3d$ with vortex filament initial data of arbitrary circulation, that is, initial vorticity given by a divergence-free vector-valued measure of arbitrary mass supported on a smooth curve. First, we prove global well-posedness for perturbations of the Oseen vortex column in scaling-critical spaces. Second, we prove local well-posedness (in a sense to be made precise) when the filament is a smooth, closed, non-self-intersecting curve. Besides their physical interest, these results are the first to give well-posedness in a neighborhood of large self-similar solutions of $3d$ Navier-Stokes, as well as solutions which are locally approximately self-similar.
△ Less
Submitted 16 May, 2022; v1 submitted 11 September, 2018;
originally announced September 2018.
-
Multiview Boosting by Controlling the Diversity and the Accuracy of View-specific Voters
Authors:
Anil Goyal,
Emilie Morvant,
Pascal Germain,
Massih-Reza Amini
Abstract:
In this paper we propose a boosting based multiview learning algorithm, referred to as PB-MVBoost, which iteratively learns i) weights over view-specific voters capturing view-specific information; and ii) weights over views by optimizing a PAC-Bayes multiview C-Bound that takes into account the accuracy of view-specific classifiers and the diversity between the views. We derive a generalization b…
▽ More
In this paper we propose a boosting based multiview learning algorithm, referred to as PB-MVBoost, which iteratively learns i) weights over view-specific voters capturing view-specific information; and ii) weights over views by optimizing a PAC-Bayes multiview C-Bound that takes into account the accuracy of view-specific classifiers and the diversity between the views. We derive a generalization bound for this strategy following the PAC-Bayes theory which is a suitable tool to deal with models expressed as weighted combination over a set of voters. Different experiments on three publicly available datasets show the efficiency of the proposed approach with respect to state-of-art models.
△ Less
Submitted 27 August, 2018; v1 submitted 17 August, 2018;
originally announced August 2018.
-
Existence and uniqueness of solutions for a quasilinear KdV equation with degenerate dispersion
Authors:
Pierre Germain,
Benjamin Harrop-Griffiths,
Jeremy Marzuola
Abstract:
We consider a quasilinear KdV equation that admits compactly supported traveling wave solutions (compactons). This model is one of the most straightforward instances of degenerate dispersion, a phenomenon that appears in a variety of physical settings as diverse as sedimentation, magma dynamics and shallow water waves. We prove the existence and uniqueness of solutions with sufficiently smooth, sp…
▽ More
We consider a quasilinear KdV equation that admits compactly supported traveling wave solutions (compactons). This model is one of the most straightforward instances of degenerate dispersion, a phenomenon that appears in a variety of physical settings as diverse as sedimentation, magma dynamics and shallow water waves. We prove the existence and uniqueness of solutions with sufficiently smooth, spatially localized initial data.
△ Less
Submitted 1 January, 2018;
originally announced January 2018.
-
Stability of the Couette flow at high Reynolds number in 2D and 3D
Authors:
Jacob Bedrossian,
Pierre Germain,
Nader Masmoudi
Abstract:
We review works on the asymptotic stability of the Couette flow. The majority of the paper is aimed towards a wide range of applied mathematicians with an additional section aimed towards experts in the mathematical analysis of PDEs.
We review works on the asymptotic stability of the Couette flow. The majority of the paper is aimed towards a wide range of applied mathematicians with an additional section aimed towards experts in the mathematical analysis of PDEs.
△ Less
Submitted 7 December, 2017;
originally announced December 2017.
-
Optimal local well-posedness theory for the kinetic wave equation
Authors:
Pierre Germain,
Alexandru D. Ionescu,
Minh-Binh Tran
Abstract:
We prove local existence and uniqueness results for the (space-homogeneous) 4-wave kinetic equation in wave turbulence theory. We consider collision operators defined by radial, but general dispersion relations satisfying suitable bounds, and we prove two local well-posedness theorems in nearly critical weighted spaces.
We prove local existence and uniqueness results for the (space-homogeneous) 4-wave kinetic equation in wave turbulence theory. We consider collision operators defined by radial, but general dispersion relations satisfying suitable bounds, and we prove two local well-posedness theorems in nearly critical weighted spaces.
△ Less
Submitted 18 January, 2018; v1 submitted 15 November, 2017;
originally announced November 2017.
-
Compactons and their variational properties for degenerate KdV and NLS in dimension 1
Authors:
Pierre Germain,
Benjamin Harrop-Griffiths,
Jeremy L. Marzuola
Abstract:
We analyze the stationary and traveling wave solutions to a family of degenerate dispersive equations of KdV and NLS-type. In stark contrast to the standard soliton solutions for non-degenerate KdV and NLS equations, the degeneracy of the elliptic operators studied here allows for compactly supported steady or traveling states. As we work in $1$ dimension, ODE methods apply, however the models con…
▽ More
We analyze the stationary and traveling wave solutions to a family of degenerate dispersive equations of KdV and NLS-type. In stark contrast to the standard soliton solutions for non-degenerate KdV and NLS equations, the degeneracy of the elliptic operators studied here allows for compactly supported steady or traveling states. As we work in $1$ dimension, ODE methods apply, however the models considered have formally conserved Hamiltonian, Mass and Momentum functionals, which allow for variational analysis as well.
△ Less
Submitted 14 September, 2017;
originally announced September 2017.
-
On the Cubic Lowest Landau Level Equation
Authors:
Patrick Gérard,
Pierre Germain,
Laurent Thomann
Abstract:
We study dynamical properties of the cubic lowest Landau level equation, which is used in the modeling of fast rotating Bose-Einstein condensates. We obtain bounds on the decay of general stationary solutions. We then provide a classification of stationary waves with a finite number of zeros. Finally, we are able to establish which of these stationary waves are stable, through a variational an…
▽ More
We study dynamical properties of the cubic lowest Landau level equation, which is used in the modeling of fast rotating Bose-Einstein condensates. We obtain bounds on the decay of general stationary solutions. We then provide a classification of stationary waves with a finite number of zeros. Finally, we are able to establish which of these stationary waves are stable, through a variational analysis.
△ Less
Submitted 13 September, 2017;
originally announced September 2017.